CN114998751A - PolSAR image ship detection joint optimization method and system - Google Patents

PolSAR image ship detection joint optimization method and system Download PDF

Info

Publication number
CN114998751A
CN114998751A CN202210927504.2A CN202210927504A CN114998751A CN 114998751 A CN114998751 A CN 114998751A CN 202210927504 A CN202210927504 A CN 202210927504A CN 114998751 A CN114998751 A CN 114998751A
Authority
CN
China
Prior art keywords
polarization
detector
joint
detection
detectors
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210927504.2A
Other languages
Chinese (zh)
Inventor
刘涛
杨子渊
方璐
高贵
刘维建
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Naval University of Engineering PLA
Original Assignee
Naval University of Engineering PLA
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Naval University of Engineering PLA filed Critical Naval University of Engineering PLA
Priority to CN202210927504.2A priority Critical patent/CN114998751A/en
Publication of CN114998751A publication Critical patent/CN114998751A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Abstract

The invention discloses a PolSAR image ship detection joint optimization method and a system, wherein prior information is used for detection on a training set by using different polarization detectors, an improved supervised classifier LDA algorithm is used for forming a joint polarization detector JD-ALL, and on the basis, in order to save hardware cost and calculation cost in actual satellite application, the invention combines the original single polarization detector results two by two, selects main contributors of the joint polarization detector JD-ALL and forms a two-joint polarization monitor JD-2. And on the basis, performing depth optimization on the DJD detector by adopting a genetic algorithm to obtain a depth joint detector DJD detector. The depth joint detector DJD has better detection performance than the joint polarization detector JD-ALL. Finally, the depth joint detector DJD obtained through optimization is used for effectively detecting the PolSAR image complex sea condition area and capturing the weak and small sea surface target.

Description

PolSAR image ship detection joint optimization method and system
Technical Field
The invention relates to the technical field of geospatial information systems, in particular to a PolSAR image ship detection joint optimization method and system.
Background
The polarimetric SAR system formed by applying the polarization information to the synthetic aperture radar SAR system reserves more complete target electromagnetic scattering characteristics, becomes an important tool of the modern radar imaging technology, and is widely applied to the fields of military reconnaissance, topographic mapping, environmental and natural disaster monitoring, sea surface ship target detection and the like at present. However, with the influence of the diversity of radar platform parameters and the complexity of sea condition environment, complex sea clutter modeling and estimation, slow small target detection, dense target detection and the like are still difficult problems of the current polarisar image ship target detection.
In the case of complex sea conditions, targets cannot be detected comprehensively and effectively and details cannot be kept by using a single ship detector. Therefore, how to effectively combine these different detectors to obtain comprehensive and effective information is an urgent problem to be solved.
Disclosure of Invention
In order to solve the technical problem, the invention provides a PolSAR image ship detection joint optimization method, which comprises the following steps:
step S1: analyzing the PolSAR image marked with the ship target to obtain a training set, a polarization covariance matrix C of detection pixel points in the training set and prior information, wherein the prior information is a clutter covariance matrix sigma C Sum target covariance matrix ∑ T (ii) a The detection pixel points in the training set comprise target pixel points and clutter pixel points;
step S2: constructing N different polarization detectors according to the prior information, applying the polarization detectors to the training set, processing the target pixel points and the clutter pixel points, taking the result of each polarization detector as a one-dimensional characteristic, and jointly generating a high-dimensional target data set X T Sum clutter dataset X C ；
Step S3: constructing a joint polarization detector JD-ALL by using an improved LDA algorithm of a supervision classifier, and drawing a detection performance ROC curve;
step S4: combining N polarization detectors pairwise to construct X dual polarization detectors JD-2, drawing a detection performance ROC curve, and comparing and analyzing the detection performance ROC curve of the X dual polarization detectors JD-2 and the detection performance ROC curve of the combined polarization detectors JD-ALL to obtain an optimal dual polarization detector JD-2 with the performance closest to that of the combined polarization detectors JD-ALL;
step S5: and optimizing the optimal combined polarization monitor JD-2 to obtain a depth combined detector DJD, and detecting the PolSAR image to be detected by using the depth combined detector DJD to obtain a detection result.
Preferably, when the target pixel point and the clutter pixel point of the training set in the step S1 are insufficient, the training set is supplemented by a monte carlo simulation mode based on the prior information.
Preferably, step S3 includes the steps of:
step S31: the projection matrix G corresponding to the joint polarization detector JD-ALL is calculated by a modified supervised classifier LDA algorithm, which is shown as follows:
Figure 290461DEST_PATH_IMAGE001
in the formula, R T 、R C Are each X T 、X C The cross-correlation matrix of (a); r SC 、R TC Are each X T 、X C Beta represents a balance factor, G H G = I represents regularization constraint, and tr represents a matrix tracing function;
step S32: calculating a detection result of the joint polarization detector JD-ALL through a polarization detection formula z = tr (GC), wherein z represents an output result of the joint polarization detector JD-ALL, C represents a polarization covariance matrix of a detection pixel point, and tr represents a matrix tracing function;
step S33: and adjusting the threshold from low to high according to the detection result, and drawing an ROC curve of the detection performance of the joint polarization detector JD-ALL.
Preferably, step S4 includes the steps of:
step S41: combining N polarization detectors pairwise to obtain X dual polarization monitors JD-2, and drawing X detection performance ROC curves corresponding to the dual polarization monitors JD-2 by the method of step S3;
step S42: and comparing the detection performance ROC curves of the X dual polarization detectors JD-2 with the detection performance ROC curves of the dual polarization detectors JD-ALL, and selecting the closest dual polarization detector JD-2 as the optimal dual polarization detector JD-2.
Preferably, in step S42, the method for comparing the detection performance ROC curve includes: and selecting a detection performance ROC curve positioned above.
Preferably, step S5 includes the steps of:
step S51: setting reference coefficients alpha and eta of two polarization detectors in the optimal two joint polarization monitor JD-2;
step S52: and (3) optimizing the alpha and eta parameters through a genetic algorithm to construct a depth joint detector DJD.
Preferably, step S52 includes the steps of:
step S521: setting a defined relation between alpha and eta:
α²+η²=1
α：η=1：-（max(b)+min(b)）/2
in the formula, b represents
Figure 161465DEST_PATH_IMAGE002
A characteristic value of (d);
step S522: determining transformation matrixes A and B corresponding to two polarization detectors respectively through the optimal dual polarization monitor JD-2, and calculating clutter output values corresponding to the clutter data sets and target output values corresponding to the target data sets through a formula z = tr (GC);
step S523: determining a threshold range through the range of the clutter output value, increasing the threshold to obtain an ROC curve, and taking the maximum AUC of the area under the ROC curve as an optimization target;
step S524: and constructing a depth joint detector DJD by a genetic algorithm based on the optimization target and the optimization parameters alpha and eta, wherein a transformation matrix P = alpha A + eta B of the depth joint detector DJD.
The invention also provides a PolSAR image ship detection combined optimization system, which comprises an information acquisition module, a polarization detector construction module, a polarization detector combined module, a performance analysis module, a depth optimization module and a detection module;
the information acquisition module is used for analyzing the PolSAR image, constructing a training set, a polarization covariance matrix and prior information of detection pixel points in the training set, and inputting the polarization covariance matrix and the prior information into the polarization detector construction module, wherein the detection pixel points in the training set comprise target pixel points and clutter pixel points;
the polarization detector constructing module is used for constructing N different polarization detectors and applying the polarization detectors to the training set to obtain a target data set and a clutter data set;
the polarization detector combination module: the method is used for constructing a joint polarization detector JD-ALL by utilizing an improved supervised classifier LDA method, and constructing X joint polarization detectors JD-2 by mutually combining the N polarization detectors pairwise;
the performance analysis module is used for analyzing the performances of the JD-ALL and the X dual polarization monitors JD-2 and selecting the optimal JD-2 which is closest to the performance of the JD-ALL;
the depth optimization module is used for optimizing the optimal combined polarization monitor JD-2 by using a genetic algorithm to construct a depth combined detector DJD;
and the detection module is used for carrying out ship target detection on the PolSAR image complex sea state area through a depth joint detector DJD to obtain a detection result.
Preferably, in the information acquisition module, when the target pixel points and the clutter pixel points of the training set are insufficient, the training set is supplemented in a monte carlo simulation mode based on the prior information.
Preferably, the depth optimization module determines a limit relationship by setting coefficients α and η of two polarization detectors in the optimal combined polarization monitor JD-2, and constructs the depth combined detector DJD by optimizing parameters α and η through a genetic algorithm.
Compared with the closest prior art, the technical scheme provided by the invention has the following beneficial effects: an improved supervised classifier LDA approach seeks a spatial transformation in a manner that maximizes inter-class separation, minimizes intra-class separation, or minimizes dissimilarity. The improved supervised classifier LDA projects the high-dimensional pattern samples to the optimal discrimination vector space so as to achieve the effects of extracting classification information and compressing the dimension of the feature space, and the maximum inter-class distance and the minimum intra-class distance of the pattern samples in a new subspace are ensured after projection, namely the pattern has the optimal separability in the space.
On the basis, the joint polarization detector JD-ALL formed by the whole combination is used as a contrast quantity, and different polarization detectors are combined pairwise to find a main contribution combination, so that the calculation quantity in the actual PolSAR image processing can be reduced. After the optimal dual combined polarization monitor JD-2 of the main contribution combination is found, the coefficients of the optimal dual combined polarization monitor JD-2 are subjected to deep optimization, the coefficients obtained by the supervisory classifier LDA are used as initial values, optimization is carried out through a genetic algorithm, breakthrough of detection performance can be further obtained on the basis of the advantages of the dual combined polarization monitors, and a deep combined detector DJD is formed. The depth joint detector DJD can perform well under complex sea conditions. The performance is greatly improved on the basis of the sea surface detection problem of the satellite-borne polarimetric synthetic aperture radar PolSAR, and the method has obvious civil and military values.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a PolSAR image of a sea area of a Radarsat-2 satellite;
FIG. 3 is a diagram of the detection result of the to-be-detected region in FIG. 2 extracted by DJD;
FIG. 4 is a ROC plot of detection performance in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
As shown in fig. 1, a flow diagram of a method for jointly optimizing PolSAR image ship detection provided by the present invention includes the following steps:
step S1: as shown in fig. 2, a polarimetric sar image of a certain sea area of a Radarsat-2 satellite is represented, the polarimetric sar image is analyzed, a polarization covariance matrix C of detection pixel points in a training set and the training set and prior information are obtained, the prior information includes a clutter covariance matrix Σ C and a target covariance matrix Σ T, and the detection pixel points in the training set include target pixel points and clutter pixel points.
In the embodiment of the invention, when the training set samples are insufficient, the targets and the clutter samples in the original training set are supplemented in a Monte Carlo simulation mode based on the prior information so as to perfect the training set and ensure that the training result is more accurate.
Step S2: with a priori information, N different polarization detectors are constructed, such as PDOF, APDOF, EVD, PWF, PMF, etc., and appliedProcessing target pixel points and clutter pixel points in a training set, taking the result of each polarization detector as a one-dimensional characteristic, and jointly generating a high-dimensional target data set X T Clutter data set X C ；
Target data set X T Clutter data set X C Each sample of the method is a high-dimensional vector, the dimension number is equal to the number N of single polarization detectors, and the existing training set is converted into a data set, so that the method is convenient for subsequent operation and reduces the operation consumption.
Step S3: constructing a joint polarization detector JD-ALL by using an improved LDA algorithm of a supervision classifier, and drawing a detection performance ROC curve;
the projection matrix G corresponding to the joint polarization detector JD-ALL is calculated by a modified LDA algorithm, G being a hermitian matrix, which can be understood as a conjugate multiplication of a certain linear transformation with itself, and the specific algorithm is as follows:
Figure 361502DEST_PATH_IMAGE003
in the formula, R T 、R C Are each X T 、X C The cross-correlation matrix of (a); r SC 、R TC Are each X T 、X C B denotes a balance factor, G H G = I denotes the regularization constraint.
The constructed joint polarization detector JD-ALL has subspace projection performance, can be projected to a maximum signal-to-clutter ratio subspace, and further improves the target detection performance.
And calculating the detection result of the joint polarization detector JD-ALL through a polarization detection formula z = tr (GC), wherein z represents the output result of the JD-ALL, tr represents a matrix tracing function, and according to the detection result, adjusting a threshold from low to high and drawing a ROC curve of the detection performance of the joint polarization detector JD-ALL.
Step S4: and combining the N polarization detectors pairwise to each other to construct X dual polarization monitors JD-2, drawing corresponding detection performance ROC curves, and analyzing to obtain the optimal dual polarization monitor JD-2 with the performance closest to that of the combined polarization detector JD-ALL, so that main contributors in the combined polarization detector JD-ALL can be extracted.
Specifically, combining N polarization detectors pairwise, and repeating step S3 to obtain X combined polarization detectors JD-2 and X corresponding detection performance ROC curves; comparing the detection performance ROC curves of the X dual polarization monitors JD-2 with the detection performance ROC curves of the combined polarization detectors JD-ALL to obtain the optimal dual polarization monitors JD-2, wherein the comparison method of the ROC curves comprises the following steps: the ROC curve located above has good performance.
FIG. 4 shows, with the exception of the uppermost curve, a ROC curve for the detection performance of nine dual polarization detectors JD-2 according to the embodiment of the present invention;
the figure is explained in turn from bottom to top:
the first is a detection performance ROC curve of the combined polarization detector RS;
the second is a detection performance ROC curve of the combined polarization detector PNF;
the third is a detection performance ROC curve of the combined polarization detector OPD and the combined polarization detector PWF;
the fourth is a detection performance ROC curve of the Joint polarization detector Joint-1, the Joint polarization detector Joint-2, the Joint polarization detector Joint-3 and the Joint polarization detector Joint-4;
the fifth line is a detection performance ROC curve of the combined polarization detector SPDOF-APDOF;
the fourth curve and the fifth curve are partially overlapped, and the fifth curve is slightly higher than the fourth curve;
it can be seen that the performance of the two joint polarization detectors spdf-apdf in the embodiment of the present invention is the most superior, and the monitoring performance closest to that of the joint polarization detector JD-ALL is the main contributor in the embodiment, so the embodiment of the present invention selects the spdf-apdf as the optimal joint polarization detector JD-2.
Step S5: and optimizing the optimal combined polarization monitor JD-2 to obtain a depth combined detector DJD, and detecting the PolSAR image by using the depth combined detector DJD to obtain a detection result.
Step S51: setting reference coefficients alpha and eta of two polarization detectors in the optimal two joint polarization monitor JD-2;
step S52: and optimizing the alpha and eta parameters through a genetic algorithm to obtain the DJD detector.
Step S521: setting a defined relation between alpha and eta:
α²+η²=1
α：η=1：-（max(b)+min(b)）/2
in the formula, b represents
Figure 777702DEST_PATH_IMAGE004
A characteristic value of (d);
step S522: determining transformation matrixes A and B corresponding to two polarization detectors respectively by an optimal dual polarization monitor JD-2, and calculating clutter output values corresponding to the clutter data set and target output values corresponding to the target data set by a formula z = tr (GC)
Step S523: determining a threshold range through the range of the clutter output value, increasing the threshold, drawing an ROC curve, and optimizing the target by taking the maximum AUC of the area under the curve;
step S524: and obtaining a transformation matrix P = alpha A + eta B of the depth joint detector DJD through a genetic algorithm based on the optimization target optimization parameters alpha and eta.
As shown in fig. 3, the detection result graph extracted by the depth joint detector DJD is a graph, the crosses represent false alarms, and the other patterns represent ships, so that it can be seen that the ship target detection is performed on the complex sea state area of the POlSAR image by the DJD detector, 11 targets are all extracted, and small targets which are difficult to discover in the original data set are discovered, and the obtained result is accurate and the performance is superior.
The invention also provides a PolSAR image ship detection combined optimization system, which comprises an information acquisition module, a polarization detector construction module, a polarization detector combined module, a performance analysis module, a depth optimization module and a detection module;
the information acquisition module is used for analyzing the PolSAR image, constructing a training set, a polarization covariance matrix and prior information of detection pixel points in the training set, and inputting the polarization covariance matrix and the prior information into the polarization detector construction module, wherein the detection pixel points in the training set comprise target pixel points and clutter pixel points;
the polarization detector constructing module is used for constructing N different polarization detectors and applying the polarization detectors to the training set to obtain a target data set and a clutter data set;
the polarization detector combination module: the method is used for constructing a joint polarization detector JD-ALL by utilizing an improved supervised classifier LDA method, and constructing X joint polarization detectors JD-2 by mutually combining the N polarization detectors pairwise;
the performance analysis module is used for analyzing the performances of the joint polarization detector JD-ALL and the X joint polarization detectors JD-2 and selecting the optimal joint polarization detector JD-2 with the performance closest to that of the joint polarization detector JD-ALL;
the depth optimization module is used for optimizing the optimal combined polarization monitor JD-2 by using a genetic algorithm to construct a depth combined detector DJD;
and the detection module is used for carrying out ship target detection on the PolSAR image complex sea state area through a depth joint detector DJD to obtain a detection result.
Preferably, in the information acquisition module, when the target pixel points and the clutter pixel points of the training set are insufficient, the training set is supplemented in a monte carlo simulation mode based on the prior information.
Preferably, the depth optimization module determines a limit relationship by setting coefficients α and η of two polarization detectors in the optimal combined polarization monitor JD-2, and constructs the depth combined detector DJD by optimizing parameters α and η through a genetic algorithm.
In order to simplify the description, all possible combinations of the above features in the above embodiments are not described, but only preferred embodiments of the present invention are shown, which are described in detail and are not to be construed as limiting the scope of the present invention. The combination of these features should be considered as the scope of the present specification unless there is any contradiction.
It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (10)

1. A PolSAR image ship detection joint optimization method is characterized by comprising the following steps: the method comprises the following steps:
step S1: analyzing the PolSAR image marked with the ship target to obtain a training set, a polarization covariance matrix C of detection pixel points in the training set and prior information, wherein the prior information is a clutter covariance matrix sigma C Sum target covariance matrix ∑ T (ii) a The detection pixel points in the training set comprise target pixel points and clutter pixel points;
step S2: constructing N different polarization detectors according to the prior information, applying the polarization detectors to the training set, processing the target pixel points and the clutter pixel points, taking the result of each polarization detector as a one-dimensional characteristic, and jointly generating a high-dimensional target data set X T Sum clutter dataset X C ；
Step S3: constructing a joint polarization detector JD-ALL by using an improved LDA algorithm of a supervision classifier, and drawing a detection performance ROC curve;
step S4: combining N polarization detectors pairwise to each other, constructing X dual polarization monitors JD-2, drawing a detection performance ROC curve, and performing comparative analysis on the detection performance ROC curve of the X dual polarization monitors JD-2 and the detection performance ROC curve of the combined polarization detectors JD-ALL to obtain an optimal second combined polarization monitor JD-2 with the performance closest to the performance of the combined polarization detectors JD-ALL;
step S5: and optimizing the optimal combined polarization monitor JD-2 to obtain a depth combined detector DJD, and detecting the PolSAR image to be detected by using the depth combined detector DJD to obtain a detection result.
2. The PolSAR image ship detection joint optimization method according to claim 1, characterized in that: and when the target pixel points and the clutter pixel points of the training set in the step 1 are insufficient, supplementing the training set by a Monte Carlo simulation mode based on the prior information.
3. The PolSAR image ship detection joint optimization method according to claim 1, characterized in that: step S3 includes the following steps:
step S31: the projection matrix G corresponding to the joint polarization detector JD-ALL is calculated by a modified supervised classifier LDA algorithm, which is shown as follows:
Figure 330952DEST_PATH_IMAGE001
in the formula, R T 、R C Are each X T 、X C The cross-correlation matrix of (a); r SC 、R TC Are each X T 、X C Beta represents a balance factor, G H G = I represents regularization constraint, and tr represents a matrix tracing function;
step S32: calculating a detection result of the joint polarization detector JD-ALL through a polarization detection formula z = tr (GC), wherein z represents an output result of the joint polarization detector JD-ALL, C represents a polarization covariance matrix of a detection pixel point, and tr represents a matrix tracing function;
step S33: and adjusting the threshold from low to high according to the detection result, and drawing an ROC curve of the detection performance of the joint polarization detector JD-ALL.
4. The PolSAR image ship detection joint optimization method according to claim 1, characterized in that: step S4 includes the following steps:
step S41: combining N polarization detectors pairwise to obtain X dual polarization monitors JD-2, and drawing X detection performance ROC curves corresponding to the dual polarization monitors JD-2 by the method of step S3;
step S42: and comparing the detection performance ROC curves of the X dual polarization detectors JD-2 with the detection performance ROC curves of the dual polarization detectors JD-ALL, and selecting the closest dual polarization detector JD-2 as the optimal dual polarization detector JD-2.
5. The PolSAR image ship detection joint optimization method according to claim 4, characterized in that: in step S42, the method for comparing the detection performance ROC curves includes: and selecting a detection performance ROC curve positioned above.
6. The PolSAR image ship detection joint optimization method according to claim 1, characterized in that: step S5 includes the following steps:
step S51: setting reference coefficients alpha and eta of two polarization detectors in the optimal two joint polarization monitor JD-2;
step S52: and (3) optimizing the alpha and eta parameters through a genetic algorithm to construct a depth joint detector DJD.
7. The PolSAR image ship detection joint optimization method of claim 6, characterized by: step S52 includes the following steps:
step S521: setting a defined relationship of α and η:
α²+η²=1
α：η=1：-（max(b)+min(b)）/2
in the formula, b represents
Figure 599122DEST_PATH_IMAGE002
A characteristic value of (d);
step S522: determining transformation matrixes A and B corresponding to two polarization detectors respectively through the optimal dual polarization monitor JD-2, and calculating clutter output values corresponding to the clutter data sets and target output values corresponding to the target data sets through a formula z = tr (GC);
step S523: determining a threshold range through the range of the clutter output value, increasing the threshold to obtain an ROC curve, and taking the maximum AUC of the area under the ROC curve as an optimization target;
step S524: and constructing a depth joint detector DJD by a genetic algorithm based on the optimization target and the optimization parameters alpha and eta, wherein a transformation matrix P = alpha A + eta B of the depth joint detector DJD.
8. A PolSAR image ship detection joint optimization system is characterized in that: the device comprises an information acquisition module, a polarization detector construction module, a polarization detector combination module, a performance analysis module, a depth optimization module and a detection module;
the information acquisition module is used for analyzing the PolSAR image, constructing a training set, detecting a polarization covariance matrix and prior information of pixel points, and inputting the polarization covariance matrix and the prior information into the polarization detector construction module;
the polarization detector constructing module is used for constructing N different polarization detectors and applying the polarization detectors to the training set to obtain a target data set and a clutter data set;
the polarization detector combination module: the method is used for constructing a joint polarization detector JD-ALL by utilizing an improved supervised classifier LDA method, and constructing X joint polarization detectors JD-2 by mutually combining the N polarization detectors pairwise;
the performance analysis module is used for analyzing the performances of the JD-ALL and the X dual polarization monitors JD-2 and selecting the optimal JD-2 which is closest to the performance of the JD-ALL;
the depth optimization module is used for optimizing the optimal combined polarization monitor JD-2 by using a genetic algorithm to construct a depth combined detector DJD;
and the detection module is used for carrying out ship target detection on the PolSAR image complex sea state area through a depth joint detector DJD to obtain a detection result.
9. The PolSAR image ship detection joint optimization system of claim 8, characterized in that: in the information acquisition module, when the target pixel points and the clutter pixel points of the training set are insufficient, based on the prior information, the training set is supplemented in a Monte Carlo simulation mode.
10. The PolSAR image ship detection joint optimization system of claim 8, wherein: and the depth optimization module determines a limit relation by setting coefficients alpha and eta of two polarization detectors in the optimal combined polarization monitor JD-2, and optimizes parameters alpha and eta by a genetic algorithm to construct a depth combined detector DJD.
CN202210927504.2A 2022-08-03 2022-08-03 PolSAR image ship detection joint optimization method and system Pending CN114998751A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210927504.2A CN114998751A (en) 2022-08-03 2022-08-03 PolSAR image ship detection joint optimization method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210927504.2A CN114998751A (en) 2022-08-03 2022-08-03 PolSAR image ship detection joint optimization method and system

Publications (1)

Publication Number Publication Date
CN114998751A true CN114998751A (en) 2022-09-02

Family

ID=83021889

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210927504.2A Pending CN114998751A (en) 2022-08-03 2022-08-03 PolSAR image ship detection joint optimization method and system

Country Status (1)

Country Link
CN (1) CN114998751A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104091335A (en) * 2014-07-04 2014-10-08 西安电子科技大学 Polarization SAR image ship target detection method
CN104239895A (en) * 2014-09-03 2014-12-24 西安电子科技大学 SAR target identification method based on feature dimension reduction
CN106291550A (en) * 2016-07-27 2017-01-04 西安电子科技大学 The polarization SAR Ship Detection of core is returned based on local scattering mechanism difference
CN113643284A (en) * 2021-09-09 2021-11-12 西南交通大学 Polarimetric synthetic aperture radar image ship detection method based on convolutional neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104091335A (en) * 2014-07-04 2014-10-08 西安电子科技大学 Polarization SAR image ship target detection method
CN104239895A (en) * 2014-09-03 2014-12-24 西安电子科技大学 SAR target identification method based on feature dimension reduction
CN106291550A (en) * 2016-07-27 2017-01-04 西安电子科技大学 The polarization SAR Ship Detection of core is returned based on local scattering mechanism difference
CN113643284A (en) * 2021-09-09 2021-11-12 西南交通大学 Polarimetric synthetic aperture radar image ship detection method based on convolutional neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
TAO LIU ET AL.: "Joint Polarimetric Subspace Detector Based on Modified Linear Discriminant Analysis", 《 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 *

Similar Documents

Publication Publication Date Title
CN111860612B (en) Unsupervised hyperspectral image hidden low-rank projection learning feature extraction method
Pei et al. SAR automatic target recognition based on multiview deep learning framework
CN109636742B (en) Mode conversion method of SAR image and visible light image based on countermeasure generation network
Matteoli et al. A tutorial overview of anomaly detection in hyperspectral images
Zhao et al. Band-subset-based clustering and fusion for hyperspectral imagery classification
Zhang et al. An efficient machine learning approach for indoor localization
CN103955701B (en) Multi-level-combined multi-look synthetic aperture radar image target recognition method
Liu et al. Track infrared point targets based on projection coefficient templates and non-linear correlation combined with Kalman prediction
CN103258324B (en) Based on the method for detecting change of remote sensing image that controlled kernel regression and super-pixel are split
CN103500450A (en) Multi-spectrum remote sensing image change detection method
Ai et al. A fine PolSAR terrain classification algorithm using the texture feature fusion-based improved convolutional autoencoder
CN103870836A (en) POCS (Projections Onto Convex Sets) super-resolution reconstruction-based SAR (Synthetic Aperture Radar) image target recognition method
Singh et al. Unsupervised change detection from remote sensing images using hybrid genetic FCM
CN112329784A (en) Correlation filtering tracking method based on space-time perception and multimodal response
CN114972735A (en) Anti-occlusion moving target tracking device and method based on ROI prediction and multi-module learning
Shaw et al. Eigen-template-based HRR-ATR with multi-look and time-recursion
Yang et al. Polarimetric SAR image classification using geodesic distances and composite kernels
Gui et al. Eigenvalue statistical components-based PU-learning for PolSAR built-up areas extraction and cross-domain analysis
CN102819838B (en) Hyperspectral remote sensing image change detection method based on multisource target characteristic support
Celik et al. Change detection without difference image computation based on multiobjective cost function optimization
Lin et al. Parameter estimation of frequency-hopping signal in uca based on deep learning and spatial time–frequency distribution
Lin et al. IR-TransDet: Infrared dim and small target detection with IR-transformer
CN114998751A (en) PolSAR image ship detection joint optimization method and system
Li et al. POLSAR Target Recognition Using a Feature Fusion Framework Based on Monogenic Signal and Complex-Valued Nonlocal Network
Xi et al. Multitarget Detection Algorithms for Multitemporal Remote Sensing Data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20220902